Data Jamboree- ggplot2

Link to paper on dataset: http://faculty.washington.edu/kenrice/heartgraphs/effectivegraphs.pdf

Dataset url: http://faculty.washington.edu/kenrice/heartgraphs/

Setup

Import packages (once per machine)

install.packages("readr")
install.packages("ggplot2")
install.packages("ggbeeswarm") 
install.packages("dplyr")
install.packages("MASS")
install.packages("hexbin")

Load packages (once per R session)

library(readr)
library(ggplot2)
library(ggbeeswarm)
library(dplyr)
library(MASS)
library(hexbin)

Import data (using readr)

• Use readr package to import our csv
• assign the data file to an R object using (<-)

heart <- read_csv("http://faculty.washington.edu/kenrice/heartgraphs/nhaneslarge.csv", na=".") #na= tells R that . is an na value

• run the command head ( ) and check with your partner to see if you got the same output

head(heart)

Univariate Plots: Folate intake by gender

Basic ggplot Anatomy

ggplot code makes graphs by layering imformation on top of an empty plot

ggplot(heart, aes(x=DR1TFOLA)) 

#This says we want ggplot to use the data.frame heart and to plot DR1TFOLA on the x-axis.

Histogram

create a histogram by adding a geom layer and using the “+” sign

ggplot(heart, aes(x=DR1TFOLA)) +
  geom_histogram()

Labels

change x-axis label from “DR1TFOLA” to “Folate Intake”

ggplot(heart, aes(x=DR1TFOLA)) +
  geom_histogram() +
  labs(x = "Folate intake") #x-axis label

Changing Colors

change outline color

ggplot(heart, aes(x=DR1TFOLA)) +
  geom_histogram(colour = "white") + 
  labs(x = "Folate intake")

change the fill of the bars

ggplot(heart, aes(x=DR1TFOLA)) +
  geom_histogram(colour = "white", fill = "peachpuff") + #yes, that is a name of a color
  labs(x = "Folate intake")

Documentation and the Help Command

language will be different for each type of graph so make sure to reference your documentation by typing “?command”

?geom_histogram

Histogram Bins and ggplot Defaults

All of our histogram plots have given us this warning:

#`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

This warning tells us that this is a default that ggplot chose for us given the data it has.

change bin size

ggplot(heart, aes(x=DR1TFOLA)) + 
  geom_histogram(colour = "white", fill = "peachpuff", bins = 50)

Facet Wrap

add facet_wrap for split plots based on a variable

ggplot(heart, aes(x = DR1TFOLA)) +
  geom_histogram(colour = "white", fill = "peachpuff", bins = 50) +
  labs(x = "Folate intake") +
  facet_wrap(~gender)

Density Plot

change the geom_ histogram layer to geom_density

ggplot(heart, aes(x = DR1TFOLA)) +
  geom_density(colour = "white", fill = "peachpuff") +  #Make sure to take out bin=50
  labs(x = "Folate intake") +
  facet_wrap(~gender)

Adding a third dimension onto one plot

plot all three variables from last plot (folate, density, and gender) on one by using color to differentiate gender

ggplot(heart, aes(x = DR1TFOLA)) +
  geom_density(aes(colour = gender)) + #Notice moving the colour into aesthetics
  labs(x = "Folate intake") 

Univariate Plots: Systolic blood pressure by gender

Dotplot - Stripchart

create a stripchart of BPXSAR(or systolic blood pressure) by gender

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_point() 

Customizing our Dotplot

add an alpha in the geom layer to make our points more transparent

• Alpha works on a scale from 0 (clear) to 1 (opaque)

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_point(alpha = .1)

add jitter to space out our points

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_jitter(alpha = .1)

Notes about Jitter:

• jitter automatically adds space (noise) to both the height and the width of your plots

• which variables won’t be effected by noice - categorical

change jitter to width only

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_jitter(alpha = .1, width = .5, height = 0)

change our x- and y-axis labels

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_jitter(alpha = .1, width = .5, height = 0) +
  labs(x = "", y = "Systolic BP (mmHg)")

Beeswarm Plot

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_beeswarm(alpha = .2) +
  labs(x = "Systolic BP (mmHg)", y = "")

Stats layers

add statistics on top of your plots with stat_summary

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_beeswarm(alpha = .2) +
  stat_summary(fun.y = "mean", geom = "point", colour = "orange") +
  labs(x = "Systolic BP (mmHg)", y = "")

Layering geom types

add other plot types as geom layers

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_boxplot(outlier.shape = NA) +   #outlier.shape=NA tells boxplot not to plot outliers
  geom_beeswarm(alpha = .2) +  
  labs(x = "Systolic BP (mmHg)", y = "")

Layer order

order is important (will change the order things are plotted)

ggplot(heart, aes(x = gender, y = BPXSAR)) +
  geom_beeswarm(alpha = .2) +  
  geom_boxplot(outlier.shape = NA) +
  labs(x = "Systolic BP (mmHg)", y = "")

Extra Challenge

Violin plot

If that was easy for you, give this a shot. Try and recreate the plots below using what you learned above.

Hint: We are using “geom_violin”.

Try adding some statistics to your plot. You can start with the mean and the median.

Try adding another geom layer.

Hint: Do you want to include your outlier points?

Bivariate Plots: Age & Systolic Blood Pressure

Simple scatterplot

ggplot(heart, aes(x = RIDAGEYR, y = BPXSAR)) +
  geom_point() +
  labs(x = "Age (years)", y = "Systolic BP (mmHg)")

If you have big n, try hexbin plot

ggplot(heart, aes(x = RIDAGEYR, y = BPXSAR)) +
  geom_hex() +
  labs(x = "Age (years)", y = "Systolic BP (mmHg)")

Add linear regression line with SE

ggplot(heart, aes(x = RIDAGEYR, y = BPXSAR)) +
  geom_point() +
  geom_smooth(method = "lm") +
  labs(x = "Age (years)", y = "Systolic BP (mmHg)")

Default is loess line

ggplot(heart, aes(x = RIDAGEYR, y = BPXSAR)) +
  geom_point() +
  geom_smooth() +
  labs(x = "Age (years)", y = "Systolic BP (mmHg)")

Add splines

library(splines)
library(MASS)
ggplot(heart, aes(x = RIDAGEYR, y = BPXSAR)) +
  geom_point() +
  stat_smooth(method = "lm", formula = y ~ ns(x, 3)) +
  labs(x = "Age (years)", y = "Systolic BP (mmHg)")

Multivariable Plots: Body Mass Index & Systolic BP, by gender and age

Just copy this:

library(dplyr)
heart2 <- heart %>% 
  mutate(age_cat = cut(RIDAGEYR,c(0,30,55,100)))

Recreate theirs first

ggplot(heart2, aes(x = BMXBMI, y = BPXSAR)) +
  geom_point() +
  stat_smooth(aes(colour = gender), method = "lm") +
  facet_wrap(~age_cat) +
  labs(x = "Body Mass Index"~(kg/m^2), y = "Systolic BP (mmHg)")

Try with facet grid, update labels

ggplot(heart2, aes(x = BMXBMI, y = BPXSAR)) +
  geom_point() +
  stat_smooth(aes(colour = gender), method = "lm") +
  facet_grid(gender~age_cat) +
  labs(x = "Body Mass Index"~(kg/m^2), y = "Systolic BP (mmHg)")

Play with colors!

ggplot(heart2, aes(x = BMXBMI, y = BPXSAR, colour = gender)) +
  geom_point(alpha = .5) +
  stat_smooth(method = "lm") +
  facet_grid(gender~age_cat) +
  theme_minimal() +
  labs(x = "Body Mass Index"~(kg/m^2), y = "Systolic BP (mmHg)") +
  scale_color_manual(values = c("#B47CC7", "#D65F5F"), guide = FALSE)

my_colors <- c("#C4AD66", "#77BEDB")
ggplot(heart2, aes(x = BMXBMI, y = BPXSAR, colour = gender)) +
  geom_point(alpha = .5) +
  stat_smooth(method = "lm") +
  facet_grid(gender~age_cat) +
  labs(x = "Body Mass Index"~(kg/m^2), y = "Systolic BP (mmHg)") +
  scale_color_manual(values = my_colors, guide = FALSE)